Slate v2 perfect plan should steal read update transaction discipline and extension DX

Decision

For the "perfect plan" target, Slate v2 should:

• steal Lexical's read/update lifecycle discipline
• steal ProseMirror's transaction and DOM-selection discipline
• steal Tiptap's extension and product DX
• reject all three engines' mismatched core identities when they fight Slate's data model and operation semantics

Steal

Lexical

• editor.read
• editor.update
• lifecycle tags
• dirty-node discipline
• extension dependency graph ideas

ProseMirror

• transaction authority
• selection mapping
• selection bookmarks
• centralized DOM selection import/export
• view-data overlay discipline

Tiptap

• extension ergonomics
• command discoverability
• selector-based React UI state
• optional chain API later

Reject

• Lexical class-based node model
• Lexical $function public style
• ProseMirror integer position model
• ProseMirror schema-first identity
• Tiptap focus().chain().run() as required public ceremony
• public mutable editor fields
• public free-transform-first docs
• React-avoidance as a performance strategy

Defer

• optional editor.chain()
• extension dependency/conflict polish
• page/layout system
• AI/review/tracked-change product surfaces
• full product UI kit parity

Public Contract

editor.read(fn)
editor.update(fn, options?)

Primitive editor methods are the power API inside editor.update.

tx.resolveTarget() stays internal.

2026-04-28 Maintain Note

The later state/tx API decision narrows this public-contract wording:

• editor.read((state) => ...) is the normal read path.
• editor.update((tx) => ...) is the normal write path.
• write groups live on tx, including plugin/extension groups.
• primitive editor.* transform methods may remain internal or advanced runtime tools, but this page should not be used to justify them as the final normal authoring DX.

Use slate-v2-state-tx-public-api-and-extension-namespaces.md as the current authority for public naming and extension namespaces.

Why This Wins

This combination gives Slate v2:

• best chance at React 19.2-perfect runtime performance
• flexible DX for custom node types
• safer browser-editing correctness than patch-driven event handling
• a migration path Plate and Yjs can actually follow

Status

Accepted as the deeper architecture/API direction for the Slate v2 perfect plan. Implementation sequencing remains a separate planning concern.